def pre_post_freqratio_comparison():
for subjID in subjIDs:
d = pd.HDFStore(os.path.join(studydir, 'Combined', '%s_silent.h5' % subjID))
df = d['df']
d.close()
df = mark_good(df)
df = df[df.good]
df['FFTabs'] = np.abs(df.FFT)
lowhighbands = dict(low=[1,8], high=[30,70])
df = EEG.calc_freqband_power_all(df, lowhighbands, fft_freq)
df['highlow'] = df.high / df.low
epochgp = df.groupby(['hemi', 'epoch'])
epochmean = epochgp.highlow.apply(np.mean)
epocherr = epochgp.highlow.apply(np.std)
colors = dict(Pre = 'b', Post = 'r')
axs = dict(l = 1, r = 2)
fig = plt.figure()
for hemi in hemis:
ax = fig.add_subplot(1, 2, axs[hemi])
ax.set_title('%s hemisphere' % hemi)
ax.bar([0, 1], epochmean[hemi][::-1], yerr = epocherr[hemi])
ax.set_xticks([0.4, 1.4])
ax.set_xticklabels(epochmean[hemi].index[::-1])
fig.suptitle('%s (%s ear lesion)' % (subjID, lesionear[subjID]))
fig.savefig(os.path.join(studydir, 'Analysis', '%s_freqratio.png' % subjID))
fig.suptitle('')
fig.clf();
def __init__(self):
QThread.__init__(self)
self.current_challenge = None
self.challenges = defaultdict(list)
self.progress = ProgressBarThread.ProgressBarThread()
self.progress.signal_progress.connect(self.change_progress)
self.progress.signal_toggle_lights.connect(self.prompter_toggle_lights)
self.progress.signal_start_new_stage.connect(self.start_new_stage)
#self.camera_thread = Camera.WebcamRecorder() # Main Camera thread
self.eeg_thread = EEG.EEG() # Main EEG client thread
self.eeg_thread.sample_signal.connect(self.get_samples)
self.mediaPlayer = QMediaPlayer(None, QMediaPlayer.StreamPlayback)
self.beeper_thread = Beeper.Beeper() # beeper thread
self.current_samples = []
self.current_original_duration = 0
self.current_eyes_open = True
self.current_movie = ""
self.used_challenges = []
self.current_state_filename = ""
self.session_labels_dict = dict()
self.symbols_count = 15 # number of repetitions for each class
self.symbols_status = defaultdict(lambda: 0) # keep track
# EEG samples collection
self.current_log_saving = False
self.current_log_filename = ""
def upload(request):
context = {}
if request.method == 'POST':
uploaded_file = request.FILES['document']
fs = FileSystemStorage()
name = fs.save(uploaded_file.name, uploaded_file)
url = fs.url(name)
context['url'] = url
data = eg.predict(pd.read_csv("/home/sjsingh/Desktop/UI/" + url))
print(data)
return render(request, 'result.html', {'data': data})
def add_lfp_filt(f, fs = 384.384384384):
lfp = f['lfp'].value
nchan, ntrials, nsamp = lfp.shape
lfp_filt = np.empty_like(lfp)
for i in range(nchan):
for j in range(ntrials):
lfp_filt[i, j, :] = EEG.remove_60hz(lfp[i, j, :])
f.create_dataset('lfp_filt', data = lfp_filt, compression = 'gzip')
def combine(epoch = 'Pre', stimtype = 'noise'):
'''
combine left and right ear stimulation blocks into one pandas DataFrame
'''
sesspaths = glob.glob(os.path.join(studydir, 'Sessions', epoch, '*'))
for sesspath in sesspaths:
absol, sessID = os.path.split(sesspath)
print sessID
fpaths = glob.glob(os.path.join(sesspath, 'fileconversion', '*.h5'))
subjIDs = np.unique([os.path.splitext(os.path.split(fpath)[1])[0].split('_')[0] for fpath in fpaths])
hemis = 'rl'
ears = 'rl'
for subjID in subjIDs:
LFP = []; LFPFILT = []; ATTEN = []; EAR = []; HEMI = []; RELHEMI = []; SESSID = [];
for ear in ears:
fpath = glob.glob(os.path.join(sesspath, 'fileconversion', '%s_%s_%s_b*.h5' % (subjID, ear.upper(), stimtype)))[0]
fin = h5py.File(fpath, 'r')
lfp = fin['lfp'].value
stimparams = fin['stimID'].value
fin.close()
nchan, ntrials, nsamp = lfp.shape
for i in xrange(nchan):
[LFP.append(lfp[i, j, :]) for j in xrange(ntrials)]
[LFPFILT.append(EEG.remove_60hz(lfp[i, j, :])) for j in xrange(ntrials)]
ATTEN.extend(stimparams[:, 1].tolist())
EAR.extend([ear]*ntrials)
HEMI.extend([hemis[i]]*ntrials)
if ear==hemis[i]:
relhemi = 'ipsi'
elif ear!=hemis[i]:
relhemi = 'contra'
RELHEMI.extend([relhemi]*ntrials)
SESSID.extend([sessID]*ntrials)
d = dict(lfp = LFP, lfpfilt = LFPFILT, atten = ATTEN, ear = EAR, hemi = HEMI, relhemi = RELHEMI, sess = SESSID)
df = pd.DataFrame(d)
fout = pd.HDFStore(os.path.join(sesspath, 'both', '%s_both.h5' % subjID))
fout['df'] = df
fout.close()
def __init__(self):
QThread.__init__(self)
self.current_challenge = None
self.challenges = defaultdict(list)
self.progress = ProgressBarThread.ProgressBarThread()
self.progress.signal_progress.connect(self.change_progress)
self.progress.signal_toggle_lights.connect(self.prompter_toggle_lights)
self.progress.signal_start_new_stage.connect(self.start_new_stage)
self.camera_thread = Camera.WebcamRecorder() # Main Camera thread
self.eeg_thread = EEG.EEG() # Main EEG client thread
self.eeg_thread.sample_signal.connect(self.get_samples)
self.mediaPlayer = QMediaPlayer(None, QMediaPlayer.StreamPlayback)
self.current_samples = []
self.current_original_duration = 0
self.current_eyes_open = True
self.current_movie = ""
self.used_challenges = []
# EEG samples collection
self.current_log_saving = False
self.current_log_filename = ""
# Training Parameters
learning_rate = 0.001
training_steps = 100
batch_size = 10
display_step = 1
# Network Parameters
num_input = 20 # data input (# signals)
# ~ timesteps = 1155 # timesteps
timesteps = 815 # timesteps
num_hidden = 128 # hidden layer num of features
num_classes = 2 # Two classes: Inter and pre
data_division = {}
patient_data = EEG.Patient_data(data_division, FLAGS.data_path)
# tf Graph input
X = tf.placeholder("float", [None, timesteps, num_input])
Y = tf.placeholder("float", [None, num_classes])
def RNN(x):
# GPU version
# ~ lstm_cuda = tf.contrib.cudnn_rnn.CudnnLSTM(1,num_hidden)
# ~ outputs, _ = lstm_cuda(x)
lstm_cell = tf.contrib.rnn.LSTMBlockCell(num_hidden, forget_bias=1.0)
# ~ lstm_cell = tf.contrib.rnn.LSTMCell(num_hidden, forget_bias=1.0)
outputs, _ = tf.nn.dynamic_rnn(cell=lstm_cell, inputs=x, dtype=tf.float32)
# we take only the output at the last time
# ~ fc_layer = tf.layers.dense(outputs[:,-1,:], 30, activation=tf.nn.relu)
import numpy as np
#import pygtk
import EEG
#p = EEG.EEG('Dog_2', 'interictal', 17)
#p = EEG.EEG('Dog_2', 'preictal', 17)
#p = EEG.EEG('Patient_2', 'interictal', 17)
p = EEG.EEG('Patient_2', 'preictal', 17)
p.load()
print p
#p.normalize_channels()
p.normalize_overall()
#print np.shape(p.data) == (16, ~240k)
data = p.data
eeg = np.rollaxis(data, 1)
# see p797 of matplotlib pdf : matshow() for 2d colour-map of (say) correlation matrix
#import matplotlib
#matplotlib.use('Qt4Agg') # Needs PySide - difficult to install, large
#matplotlib.use('GTKAgg') # Needs pygtk - found script on Gist.github.com
#matplotlib.use('TkAgg') # Needs Tkinter - difficult within VirtualEnv
import matplotlib.pyplot as plt
#data = p.data
n = p.n_channels
spacing = 5
trace_levels = spacing * np.arange(n, 0, -1)
import config
import FC
import TCN
import RNN
import CNN
flags = tf.flags
flags.DEFINE_string("data_path", None,
"Where the training/test data is stored.")
flags.DEFINE_string("NN", None, "Type of neural network.")
flags.DEFINE_string("patient", None, "Patient number")
FLAGS = flags.FLAGS
cfg = config.Config(data_path=FLAGS.data_path,
NN=FLAGS.NN,
patient=int(FLAGS.patient))
patient_data = EEG.Patient_data(cfg)
cfg.N_features = np.shape(patient_data.segments)[1]
tf.random.set_random_seed(1)
# tf Graph input
X = tf.placeholder("float", [None, 1, cfg.N_features, cfg.num_inputs])
Y = tf.placeholder("float", [None, cfg.num_classes])
if (cfg.NN == "TCN"):
logits = TCN.TCN(X, cfg)
elif (cfg.NN == "FC"):
logits = FC.FC(X, cfg)
elif (cfg.NN == "RNN"):
logits = RNN.RNN(X, cfg)
elif (cfg.NN == "CNN"):
logits = CNN.CNN(X, cfg)
f = h5py.File(fpaths[0]) nlfp = f['lfp'].value.shape[2] ustimid = np.unique(f['stimID'].value[:, 1]) print ustimid print '\t\t#LFP samples:\t%u' % nlfp print '\t\t' f.close() ''' def combine_pre_post(): for subjID in subjIDs: for epoch in epochs: print subjID, epoch sesspaths = os.path.join(studydir, 'Sessions', epoch) for sesspath in sesspaths:''' ''' EEG.fileconvert_all(experiment = 'awakeeeg', epoch = 'Pre') convert_to_pd(epoch = 'Pre') '''
import EEG as t
import pandas as pd
t.predict(pd.read_csv('/home/sjsingh/Desktop/Smart EEG/test - test.csv'))
all_sprites.add(PT1)
all_sprites.add(P1)
platforms = pygame.sprite.Group()
platforms.add(PT1)
for x in range(random.randint(4,5)):
C = True
pl = platform()
while C:
pl = platform()
C = check(pl, platforms)
platforms.add(pl)
all_sprites.add(pl)
s = EEG.SerialGet(baud=BAUD, port=PORT)
length = len(s.ypr)
while True:
s.readval()
jumpstrength = 0
P1.update()
for event in pygame.event.get():
if event.type == QUIT:
pygame.quit()
sys.exit()
for i in range(length):
if s.ypr[i] >= THRESHOLD:
to_hickle = dict(
features=all_params,
signal_period_starts=signal_period_starts,
)
# Dump data, with compression
f = "data/feat/%s/%s_%s_segment_%04d.hickle" % (p.subject, p.subject,
p.desc, p.num)
hickle.dump(to_hickle, f, mode='w', compression='gzip')
if (args.fft > 0):
if False: # Just do one sample
#p = EEG.EEG(_subject, 'interictal', 17)
p = EEG.EEG(_subject, 'preictal', 18)
preprocess(p)
exit(1)
if True: # Load in the survey, and do the fft thing for everything
#d = "data/orig/%s/" % (_subject, )
csv = open("data/survey.%s.csv" % (_subject, ), 'r')
headers = csv.readline()
for line in csv.readlines():
p = EEG.EEG(_subject, '', '') # Nonsense entries
p.survey_line_read(line)
preprocess(p)
if (args.scale > 0):
## concatinate entries (NB, must do train first, to generate min/max meta-data)
import argparse
parser = argparse.ArgumentParser(description='Survey the data')
parser.add_argument('--subject',
type=str,
required=True,
help="Dog_{1,2,3,4,5}, Patient_{1,2}")
args = parser.parse_args()
# 'Dog_2', 'Patient_2'
_subject = args.subject
d = "data/orig/%s/" % (_subject, )
csv = open("data/survey.%s.csv" % (_subject, ), 'w')
csv.write(EEG.EEG.survey_header() + "\n")
f_match = re.compile(r'%s_(.*?)_segment_(\d*?)\.mat' % (_subject, ))
for f in sorted(os.listdir(d)):
m = re.match(f_match, f)
if m is None: continue
desc, num = m.group(1), int(m.group(2))
print desc, num
p = EEG.EEG(_subject, desc, num)
p.load()
csv.write(p.survey_line_write())
csv.close()